Machine Learning Seminar: Chaffey -- Computing gradients in analog circuits

The details about the machine learning seminar are as follows: 

Time: Mon 21 Sep (2:00 - 3:00pm): 

Location: SMRI Seminar Room (A12-03-301) A12 Macleay Building, Level 3, Room 301.  

Speaker: Thomas Chaffey (USYD) 

Title: Computing gradients in analog circuits 

Abstract: With the rising costs of machine learning, analog computing has seen a new
wave of interest. Implementing neural networks in analog circuits may improve energy
efficiency and inference speed by many orders of magnitude. However, such devices
have not been able to be scaled to useful sizes due to the manufacturing variance of
analog devices. A possible solution is to train analog hardware directly, adjusting
device parameters in response to circuit measurements, using novel devices such as
memristors and other nonvolatile memories. This talk will describe ongoing research
on learning algorithms for analog electronic networks, exploiting the circuit structure
and device properties to give fast methods to perform gradient descent using hardware
measurements, with guaranteed convergence.  

Biography: Thomas Chaffey is a lecturer in the School of Electrical and Computer
Engineering at the University of Sydney, Australia. He received the B.Sc. (advmath)
degree in mathematics and computer science and the M.P.E. degree in mechanical
engineering from the University of Sydney in 2015 and 2018, respectively, and the Ph.D.
degree from the University of Cambridge, U.K., in 2022. From 2022 to 2025 he held
the Maudslay-Butler Research Fellowship in Engineering at Pembroke College, University
of Cambridge. His research interests are in nonlinear control and its intersection
with optimization, circuit theory and learning.